From 2004 to mid-2007, Iraq was extremely violent: civilian fatalities averaged more than 1,500

a month by August 2006, and by late fall, the U.S. military was suffering a

monthly toll of almost 100 dead and 700 wounded. Then something changed.

By the end of 2007, U.S. military fatalities had declined from their wartime

monthly peak of 126 in May of that year to just 23 by December. From June

2008 to June 2011, monthly U.S. military fatalities averaged fewer than 11, a

rate less than 15 percent of the 2004 through mid-2007 average and an order of

magnitude smaller than their maximum. Monthly civilian fatalities fell from

more than 1,700 in May 2007 to around 500 by December; from June 2008 to

June 2011, these averaged around 200, or about one-tenth of the rate for the last

half of 2006.1

Iraq today is far from a Garden of Eden, and a return to open warfare cannot

be ruled out. But whatever Iraq’s current politics and future prognosis, its past

now includes a remarkable reversal in 2007 from years of intense bloodshed to

almost four years of relative calm since then. What caused this turnaround?

Many analysts now credit what is commonly called “the surge” for this out-

come. On January 10, 2007, President George W. Bush announced a roughly

30,000-soldier reinforcement of the U.S. presence in Iraq, together with a new

commander in Gen. David Petraeus and a new strategy for the use of U.S.

forces. In particular, Petraeus replaced a prior emphasis on large, fortiªed

Testing the Surge

Testing the Surge Stephen Biddle,Jeffrey A. Friedman,andJacob N. Shapiro

Why Did ViolenceDecline in Iraq in 2007?

Stephen Biddle is Roger Hertog Senior Fellow for Defense Policy at the Council on Foreign Relations. JeffreyA. Friedman is a Ph.D. candidate in Public Policy at Harvard University. Jacob N. Shapiro is AssistantProfessor of Politics and International Affairs at Princeton University.

The authors thank Daniel Altman, Eli Berman, Richard Betts, Daniel Byman, Luke Condra, BenConnable, James Fearon, Peter Feaver, Gian Gentile, Carter Malkasian, Jeffrey Peterson, KennethPollack, Dan Reiter, Alissa Rubin, Idean Salehyan, Paul Staniland, Nathan Toronto, Barbara Wal-ter, Nils Weidmann, Michael Yankovich, Yuri Zhukov, and seminar participants at the Universityof California at Berkeley, George Washington University, Harvard University, the MassachusettsInstitute of Technology, Stanford University, and the 2011 PRIO-MIT Conference on Counterinsur-gency and Counterterrorism for helpful comments on earlier drafts. They also thank JoshuaBorkowski, Michael Johnson, Christopher Paik, and Ari Rubin for their excellent research assis-tance; Charles Lewis and Nathan Toronto for assistance in arranging interviews; Zeynep Bulutgilfor generously sharing her interviews on events in Ramadi in 2005–06; and the seventy U.S. andallied ofªcers whose interviews form an important part of the evidence presented below, andwithout whose time and cooperation this analysis would not have been possible. Jacob Shapiro ac-knowledges ªnancial support from Air Force Ofªce of Scientiªc Research Award #FA9550-09-1-0314.

1. Casualty ªgures are from “Operation Iraqi Freedom,” iCasualties.org, accessed July 25, 2011.

International Security, Vol. 37, No. 1 (Summer 2012), pp. 7–40© 2012 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology.

7

bases, mounted patrols, and transition to Iraqi security forces with a new pat-

tern of smaller, dispersed bases, dismounted patrolling, and direct provision of

U.S. security for threatened Iraqi civilians. Proponents of the “surge thesis”

hold that this combination of more troops and different methods reduced the

level of violence by suffocating the insurgency and destroying its ability to kill

Americans or Iraqis.2

Not everyone agrees. Critics have advanced a variety of alternative explan-

ations, including the 2006–07 Sunni tribal uprising against al-Qaida in Iraq

(AQI), which produced the “Anbar Awakening,”3 the dynamics of sectarian

cleansing,4 and interaction effects among multiple causes.5

International Security 37:1 8

2. See, for example, Kimberly Kagan, The Surge: A Military History (New York: Encounter Books,2009); John McCain and Joe Lieberman, “The Surge Worked,” Wall Street Journal, January 10, 2008;Max Boot, “The Truth about Iraq’s Casualty Count,” Wall Street Journal, May 3, 2008; James R.Crider, “A View from Inside the Surge,” Military Review, Vol. 89, No. 2 (March/April 2009), pp. 81–88; Craig A. Collier, “Now That We’re Leaving Iraq, What Did We Learn?” Military Review, Vol. 90,No. 5 (September/October 2010), pp. 88–93; Bartle Bull, “Mission Accomplished,” Prospect, Octo-ber 27, 2007; and Dale Andrade, Surging South of Baghdad: The 3d Infantry Division and Task ForceMarne in Iraq, 2007–2008 (Washington, D.C.: U.S. Army Center of Military History, 2010).3. See, for example, Austin Long, “The Anbar Awakening,” Survival, Vol. 50, No. 2 (April/May2008), pp. 67–94; Steven Simon, “The Price of the Surge,” Foreign Affairs, Vol. 87, No. 3 (May/June2008), pp. 57–76; Marc Lynch, “Sunni World,” American Prospect, September 13, 2007; Jim Michaels,A Chance in Hell: The Men Who Triumphed Over Iraq’s Deadliest City and Turned the Tide of War (NewYork: St. Martin’s, 2010); Daniel R. Green, “The Fallujah Awakening: A Case Study in Counter-Insurgency,” Small Wars and Insurgencies, Vol. 21, No. 4 (December 2010), pp. 591–609; Dick Couch,The Sheriff of Ramadi: Navy SEALs and the Winning of Anbar (Annapolis: Naval Institute Press, 2008);Stanley Kober, “Did the Surge Work?” Daily Caller, July 16, 2010; Jon Lee Anderson, “Inside theSurge: The American Military Finds New Allies, but at What Cost?” New Yorker, November 19,2009, pp. 58–69; and Nir Rosen, “The Myth of the Surge,” Rolling Stone, March 6, 2008, pp. 46–53.Following this literature, we use the phrase “Anbar Awakening” to refer both to the origins of theSunni realignment in Anbar in the fall of 2006 and to the subsequent spread and maturation of thismovement as “Sons of Iraq” (SOI) over the course of 2007. Note that, although the Awakeningthus began before the surge, the overwhelming majority of SOIs joined after April 2007, well intothe surge. See, for example, Andrade, Surging South of Baghdad, pp. 209–242.4. Nils B. Weidmann and Idean Salehyan, “Violence and Ethnic Segregation: A ComputationalModel Applied to Baghdad,” International Studies Quarterly, forthcoming; Lawrence Korb, BrianKatulis, Sean Duggan, and Peter Juul, How Does This End? Strategic Failures Overshadow TacticalGains in Iraq (Washington, D.C.: Center for American Progress, 2008); and John Agnew, Thomas W.Gillespie, Jorge Gonzalez, and Brian Min, “Baghdad Nights: Evaluating the U.S. Military ‘Surge’Using Nighttime Light Signatures,” Environment and Planning A, Vol. 40, No. 10 (October 2008),pp. 2285–2295.5. Stephen Biddle, “Stabilizing Iraq from the Bottom Up,” testimony before the U.S. Senate For-eign Relations Committee, April 2, 2008; Stephen Biddle, Michael O’Hanlon, and Kenneth Pollack,“How to Leave a Stable Iraq: Building on Progress,” Foreign Affairs, Vol. 87, No. 5 (September/October 2008), pp. 40–58; Colin H. Kahl, “Walk before Running,” Foreign Affairs, Vol. 87, No. 4(July/August 2008), pp. 151–154, which also credits congressional threats of withdrawal; andCarter Malkasian, “Did the Coalition Need More Forces in Iraq?” Joint Force Quarterly, Vol. 46,No. 3 (Summer 2007), pp. 120–126. Also emphasizing the Awakening along with the surge butwithout any explicit analysis of synergy are Bing West, The Strongest Tribe: War, Politics, and theEndgame in Iraq (New York: Random House, 2008); Linda Robinson, Tell Me How This Ends: GeneralDavid Petraeus and the Search for a Way Out of Iraq (New York: PublicAffairs, 2008); and Thomas E.Ricks, The Gamble: General David Petraeus and the American Military Adventure in Iraq, 2006–2008(New York: Penguin, 2009).

The differences among these explanations matter both for policy and for

scholarship.6 The Iraq surge is now widely seen as one of the most remarkable

military events of recent memory, and it casts a long shadow over military

doctrine and planning across much of the Western world. The British army, for

example, recently published a new counterinsurgency (COIN) doctrine shaped

by its view of the 2007 surge, and other NATO members are considering simi-

lar moves.7 Among the more important debates in U.S. defense policy today is

the appropriate balance between COIN and conventional capability; this de-

bate is powerfully inºuenced by assumptions about the role of U.S. strategy in

reversing Iraq’s violence in 2007.8 The post-2009 debate on Afghanistan has

been shaped by perceptions of the surge experience in Iraq; if these percep-

tions were unsound, much of this debate has been miscast.9 The 2007 case even

played a part in U.S. presidential politics: in the 2008 elections, John McCain’s

campaign publicly embraced a surge-only account of putative success in

Iraq, which Barack Obama’s campaign countered with an Awakening-only re-

buttal.10 Rarely do arguments over military cause and effect rise to this level of

public awareness; the 2007 campaign in Iraq has had a perhaps unique politi-

cal salience.

These debates all hinge on the surge’s relative importance in reducing Iraq’s

violence. If the commonplace surge narrative is correct, then U.S. policies were

chieºy responsible for the outcome, similar policies should work again in the

future, and defense planning should reºect this conclusion. If the Awakening

or cleansing accounts are correct, then U.S. policies had little to do with the re-

duction of violence in Iraq, future surges would be much more problematic,

and defense planning built on surge analogies would be ill advised. If the

Testing the Surge 9

6. Some observers have also argued that leadership targeting had brought the insurgency to itsknees by late 2007. See, especially, Bob Woodward, The War Within: A Secret White House History,2006–2008 (New York: Simon and Schuster, 2008). See also Christopher J. Lamb and Evan Mun-sing, Secret Weapon: High-Value Target Teams as an Organizational Innovation (Washington, D.C.:National Defense University Press, 2011). In principle, one could propose combinations other thanthe surge-Awakening thesis treated below. The explanations considered here are thus not a logi-cally exhaustive set, but the four schools we discuss capture the main lines of debate in the litera-ture to date; thus our analysis focuses on them.7. British Army Field Manual, Vol. 1, pt. 10: Countering Insurgency, Army Code 71876, October 2009.8. On the contours of this debate and how it has been inºuenced by the surge, see David H. Ucko,The New Counterinsurgency Era: Transforming the U.S. Military for Modern Wars (Washington, D.C.:Georgetown University Press, 2009), especially chap. 8. For a sharply different view, see Gian P.Gentile, “Our COIN Doctrine Removes the Enemy from the Essence of War,” Armed Forces Journal,January 2008, p. 39.9. On Iraq’s inºuence on the Afghanistan debate, see, for example, Jeffrey Michaels and MatthewFord, “Bandwagonistas: Rhetorical Re-description, Strategic Choice, and the Politics of Counterin-surgency,” Small Wars and Insurgencies, Vol. 22, No. 2 (May 2011), pp. 352–384; and Stephen Biddle,“Iraq’s Lessons in Afghanistan and Iraq,” in Seyom Brown and Robert H. Scales, eds., U.S. Policyin Afghanistan and Iraq: Lessons and Legacies (Boulder, Colo.: Lynne Rienner, 2012), pp. 89–98.10. See, for example, the transcript of the presidential debate held on September 26, 2008, http://elections.nytimes.com/2008/president/debates/transcripts/ªrst-presidential-debate.html.

surge was necessary but insufªcient without the Awakening, then U.S. policy

deserves some credit, as without it Iraqi violence would have stayed high, but

the surge would be a poor template for future policy unless similar precondi-

tions obtain elsewhere.

For scholars, too, the 2007 experience matters. A burgeoning literature in

comparative politics and international relations seeks to understand the dy-

namics of civil warfare.11 Yet, to date, little of this literature considers the role

of systematic realignments of the kind seen in the Anbar Awakening. Similarly,

the role of troop density, so important for the surge debate, remains largely ab-

sent from the scholarly literature on internal war, and the debate over partition

as a solution to ethnic conºict has yet to consider the case evidence asserted by

those who see sectarian unmixing as the reason for the reduction of violence in

Iraq.12 The U.S. military’s current COIN doctrine has been the subject of schol-

arly debate and is widely assigned for classroom use, yet its relationship to one

of the most salient recent cases of counterinsurgency—Iraq—turns on an unre-

solved question of the causes for 2007’s drop in violence.13

Yet, for all its importance, this debate has not moved from hypothesis to test.

The competing accounts emerged quickly—each was in print before the surge

had even ended. Few, however, addressed others’ claims in any depth, and

none has yet advanced a body of systematic evidence sufªcient to establish it-

self over the others on a rigorous basis. The purpose of this article is to provide

such a test by evaluating the competing hypotheses head-to-head against a

broad range of qualitative and quantitative evidence on the conduct of the

war.

This evidence suggests that a synergistic interaction between the surge and

the Awakening is the best explanation for why violence declined in Iraq in

2007. Without the surge, the Anbar Awakening would probably not have

spread fast or far enough. And without the surge, sectarian violence would

likely have continued for a long time to come—the pattern and distribution of

the bloodshed offers little reason to believe that it had burned itself out by

mid-2007. Yet the surge, though necessary, was insufªcient to explain 2007’s

sudden reversal in fortunes. Without the Awakening to thin the insurgents’

International Security 37:1 10

11. For recent reviews of the literature, see Christopher Blattman and Edward Miguel, “CivilWar,” Journal of Economic Literature, Vol. 48, No. 1 (March 2010), pp. 3–57; and Stathis N. Kalyvasand Paul Kenny, “Civil Wars,” in Robert A. Denemark, ed., The International Studies AssociationCompendium Project (Oxford: Blackwell, 2009).12. For an exception, see Jeffrey Friedman, “Manpower and Counterinsurgency: Empirical Foun-dations for Theory and Doctrine,” Security Studies, Vol. 20, No. 4 (December 2011), pp. 556–591.13. See, for example, Jeffrey C. Isaac, Stephen Biddle, Stathis Kalyvas, Wendy Brown, and DouglasA. Ollivant, “The New U.S. Army/Marine Corps Counterinsurgency Field Manual as Political Sci-ence and Political Praxis,” Perspectives on Politics, Vol. 6, No. 2 (June 2008), pp. 347–350.

ranks and unveil the holdouts to U.S. troops, the violence would probably

have remained very high until well after the surge had been withdrawn and

well after U.S. voters had lost patience with the war. Our argument is more

than just a claim that both the surge and the Awakening mattered—we argue

that a synergistic interaction between them created something new that neither

could have achieved alone. This implies that U.S. reinforcements and doctrine

played an essential role in 2007—but so did local conditions that will not nec-

essarily recur elsewhere. U.S. policy thus deserves important but partial credit

for the reduction of violence in Iraq, and similar results cannot necessarily be

expected from similar methods in the future.

We make this argument in ªve steps. First, we describe our evidence and

approach. Second, we explain why the cleansing thesis cannot account for

stability in Iraq, and how other factors must have played a critical role. We

then show that the Awakening could not have spread sufªciently without

the surge. Next we show that the surge could not have succeeded without the

Awakening, and that mutual reinforcement between these effects was neces-

sary for reducing violence so quickly and systematically across Iraq. We con-

clude by discussing our ªndings’ implications for policy and scholarship.

Evidence and Approach

We base our ªndings on two principal sources of evidence. The ªrst is a

recently declassiªed dataset of 193,264 “signiªcant activities” (SIGACTs) re-

corded by Multinational Force–Iraq (MNF-I) headquarters from February 2004

to December 2008. Each SIGACT documents a use of force involving coalition

forces, Iraqi units, insurgents, or sectarian militias reported through MNF-I

channels. These data provide the location, date, time, and nature of each

event; all such events reported to MNF-I in the time interval considered are

included.14

The second source of evidence is an original series of seventy structured in-

terviews with coalition ofªcers who fought in the 2006–08 campaign and could

observe its conduct ªrsthand. These interviews cover twenty-two of the

twenty-ªve districts responsible for 90 percent of the SIGACTs in 2006, the vio-

lence whose reduction we seek to explain. For fourteen of these districts,

we have at least two different interview subjects; for sixteen we have inter-

Testing the Surge 11

14. SIGACT data were provided by the Empirical Studies of Conºict (ESOC) Project. For a full dis-cussion, see Eli Berman, Jacob N. Shapiro, and Joseph Felter, “Can Hearts and Minds Be Bought?The Economics of Counterinsurgency in Iraq,” Journal of Political Economy, Vol. 119, No. 4 (August2011), pp. 766–819.

view coverage both before and after the violence peak in that district.15 These

interviews were conducted by the authors at Fort Leavenworth, Kansas; in

Boston, Massachusetts; and via telephone. They are documented in audio ªles

deposited at the U.S. Army Military History Institute archive in Carlisle,

Pennsylvania. The typical interview lasted 30 to 60 minutes; twenty-ªve sub-

jects granted follow-up interviews or responded to follow-up questions via

email.16

We supplement these primary sources with data on civilian casualties and

sectarian violence not caused by combat from Iraq Body Count (IBC), a non-

proªt organization dedicated to tracking civilian casualties using media re-

ports as well as hospital, morgue, and other ªgures.17 These data capture

19,961 incidents in which civilians were killed that can be accurately geo-

located to the district level, accounting for 59,245 civilian deaths from March

2003 through June 2009.18

These sources are systematic and objective, but they are not perfect.

SIGACTs undercount actual violence, because they record only episodes re-

ported to coalition authorities and then entered into a database. In addition,

SIGACTs do not measure the intensity of violent events. Participant interviews

are only as good as the accuracy of the participants’ observations and recollec-

tion. We found no evidence, however, that either source of imperfection intro-

duced systematic bias. On balance, these sources provide an unusually

objective and consistent base of information, both for tracking changes in vio-

International Security 37:1 12

15. Iraq’s violence was never uniformly distributed over the country’s 111 districts: the 25 that ac-counted for 90 percent of 2006 SIGACTs were localized in central and western Iraq, especially inAnbar, Baghdad, Diyala, and Salah ad Din Provinces. As our purpose is to explain why violencefell, the relevant explanatory universe thus consists of districts where there was violence to reduce,of which our interviews span the great majority. There is substantial variance in the scale andspeed of reduction across these 25 districts; the 22 for which we have interviews cover districtswith greater and lesser SIGACTs as well as faster and slower reductions, and are not subject to biasfrom selection on the dependent variable. Note that Iraq had 104 districts in 2004 (the start date forour violence data) but currently has 111, as several districts in the Kurdish regions have been split.16. Interviewees responded to a general call for participation to students and faculty in mid-careereducation programs at the U.S. Army Command and General Staff College, Harvard University,and Marine Corps Base Quantico. Initial participants recommended other individuals, some ofwhom were then interviewed. No selection criteria were imposed beyond service in Iraq from2006 to 2008. These student populations are large and diverse, with no reason to expect systematiccorrelation between membership and the nature of their experience in Iraq. We have no evidenceof sample bias relevant to our analysis.17. “Iraq Body Count,” http://www.iraqbodycount.org/. Our data were produced through amultiyear ESOC-IBC collaboration, which made several improvements to the publicly availableIBC data, including more consistent district-level geocoding.18. For diagnostics and a complete discussion of these data, see Luke Condra and Jacob N.Shapiro, “Who Takes the Blame? The Strategic Effects of Collateral Damage,” American Journal ofPolitical Science, Vol. 56, No. 1 (January 2012), pp. 167–187.

lence over time (which helps control for underreporting) and for identifying

changes in local political dynamics, both of which are critical for our analysis.

We use this evidence to process trace the four candidate explanations for the

Iraq case. That is, we deduce from each explanation’s causal logic a series of

observable implications that should hold if that explanation were sound. We

use those implications to guide our search for evidence and compare the evi-

dence we ªnd to what we should expect to see if that thesis were true. We then

evaluate the relative consistency of each explanation’s logical implications

with the observed evidence.19

We focus on local cross-sectional and time-series variance at the level of

individual districts or unit areas of operation (AOs) within Iraq.20 Previous

literature on the surge has often assumed that analysts’ knowledge is inher-

ently limited by the difªculties of drawing causal inference from a single

observation,21 but this problem holds only if one views Iraq as a unitary case.

In fact, there is substantial within-case variance across both space and time,

and our analysis uses geocoded, micro-level data to break the case down and

create leverage for distinguishing among the candidate explanations in the

literature.22

Sectarian Cleansing

The ªrst explanation for the reduction of violence in Iraq is that sectarian

bloodshed had played itself out by mid-2007; or as Patrick Cockburn put it,

“[T]he killing stopped because there was no one left to kill.”23 Cockburn’s

Testing the Surge 13

19. On process tracing, see Gary King, Robert Keohane, and Sidney Verba, Designing Social Inquiry:Scientiªc Inference in Qualitative Research (Princeton, N.J.: Princeton University Press, 1994), pp. 85–87, 225–228. Our approach constitutes what Alexander L. George and Andrew Bennett call “ana-lytical process tracing,” as distinct from detailed narrative description of event sequences. Georgeand Bennett, Case Studies and Theory Development in the Social Sciences (Cambridge, Mass.: MITPress, 2005), p. 211. We do not test the competing explanations in a regression framework for sev-eral reasons. First, many of the arguments have no obvious implications for variance between ob-servable factors at ªxed, comparable geographic units (e.g., district-months). Second, there is nosystematic theaterwide data on important variables such as Awakening forces’ availability. Third,and most important, there is no viable source of plausibly exogenous variation in critical variablessuch as coalition force levels or operational methods.20. Note that “AOs,” which are delineated by military formation boundaries, are rarely collin-ear with “districts,” which are Iraqi political subdivisions. We use both units of analysis asappropriate.21. Weidmann and Salehyan, “Violence and Ethnic Segregation,” p. 4; and Douglas Ollivant,Countering the New Orthodoxy: Reinterpreting Counterinsurgency in Iraq (Washington, D.C.: NewAmerica Foundation, 2011), p. 11.22. Replication data, additional details on the AOs, coding choices, and various robustness checksare provided in supplementary materials at http://esoc.princeton.edu.23. Patrick Cockburn, “Who Is Whose Enemy?” London Review of Books, Vol. 30, No. 5 (March2008), p. 14. On the cleansing school, generally, see Weidmann and Salehyan, “Violence and Ethnic

claim is obviously not accurate in literal terms: to this day, Iraq has many

Sunnis left for Shiites to kill if they were so inclined and vice versa. Instead the

cleansing thesis’s causal logic concerns the distribution of populations and sec-

tarian violence. Proponents of the cleansing thesis argue that it was the spatial

intermingling of prewar Sunnis and Shiites that led to violence: large, inter-

nally homogeneous communities would be defensible and thus secure, but the

prewar patchwork quilt of interpenetrated neighborhoods created a security

dilemma in which each group was exposed to violence from the other. In this

view, the war was chieºy a response to mutual threat, with each side ªghting

to evict rivals from areas that could then be made homogeneous and secure.

While the populations were intermingled, the violence was intense, but the

ªghting progressively unmixed the two groups, yielding large, contiguous

areas of uniform makeup with defensible borders between them. This in turn

resolved the security dilemma, and as neighborhoods were cleansed, the ªght-

ing petered out as a product of its own dynamics rather than as a response to

U.S. reinforcements.24

On its face, the cleansing thesis has major challenges to overcome. Most of

this literature advances cleansing and its burnout as an alternative to the surge

account of Iraq’s violence reduction in 2007. To sustain this central claim logi-

cally requires either that combat in areas where cleansing was happening

made up the bulk of the pre-2007 violence (hence cleansing’s completion could

end that violence), or that combat in other areas was epiphenomenal to cleans-

ing. Neither assertion is consistent with the evidence.

Throughout 2005 and 2006, much of the violence in Iraq occurred in Anbar

Province, which is almost entirely Sunni and where no unmixing could thus

occur. In fact, for most of 2006, SIGACTs data show more insurgent attacks in

International Security 37:1 14

Segregation”; Korb et al., How Does This End?; and Agnew et al., “Baghdad Nights.” This argu-ment draws on a theoretical tradition that sees security dilemmas involving comingled popula-tions as a major source of violence. See, for example, James D. Fearon and David D. Laitin,“Explaining Interethnic Cooperation,” American Political Science Review, Vol. 90, No. 4 (December1996), pp. 715–735; Barry R. Posen, “The Security Dilemma and Ethnic Conºict,” Survival, Vol. 35,No. 1 (Spring 1993), pp. 27–47; Chaim Kaufmann, “Possible and Impossible Solutions to EthnicCivil Wars,” International Security, Vol. 20, No. 4 (Spring 1996), pp. 136–175; Alexander B. Downes,“The Problem with Negotiated Settlements to Ethnic Civil Wars,” Security Studies, Vol. 13, No. 4(Summer 2004), pp. 230–279; the special issue of Security Studies, Vol. 13. No. 4 (Summer 2004); andRui J.P. de Figueiredo Jr. and Barry R. Weingast, “The Rationality of Fear: Political Opportunismand Ethnic Conºict,” in Barbara Walter and Jack Snyder, eds., Civil Wars, Insecurity, and Interven-tion (New York: Columbia University Press, 1999).24. Of course, sectarian Sunni versus Shiite killings made up only part of Iraq’s violence—it ex-cludes, for example, the Iraqi versus U.S. ªghting that loomed so large for the U.S. debate. In addi-tion, the relative prevalence of sectarian and nonsectarian violence varied both geographically andtemporally. The cleansing school is rarely explicit on what aspects of this violence it seeks toexplain.

Sunni districts than in mixed ones, and violence began to decline in Sunni

areas a full eight months before it did in mixed areas. The timing of the reduc-

tions in violence also suggests that nonsectarian violence (such as attacks on

U.S. forces) was not epiphenomenal to sectarian bloodshed. The IBC data

show that sectarian violence actually lags behind both insurgent attacks on co-

alition forces and civilian casualties resulting from combat: total SIGACTs be-

gan to decline in May 2007, three months before sectarian violence turned

around.

The cleansing thesis nevertheless enjoys a degree of intuitive appeal given

its theoretical motivation and its potential to explain not just why but when vi-

olence would decline: Iraq’s population had become much less intermingled

overall by mid-2007, and this is also when aggregate violence fell. Yet a closer

look at the violence in Baghdad offers little support for this argument: when

neighborhoods unmixed, violence moved but did not diminish, and this on-

ward advance of a moving combat frontier was far from exhausted when vio-

lence nevertheless fell in mid-2007.25

After al-Qaida in Iraq bombed the Shiite Golden Dome Mosque in Samarra

in February 2006, sectarian violence in Iraq increased, and the initial wave

did indeed focus on Shiite efforts to remove Sunnis from mixed districts in

Baghdad. The ªghting did not die out, however, when this unmixing was com-

plete. Instead, Shiite militias used the newly secure cleansed zones as bases

for onward movement into adjoining, homogeneously Sunni neighborhoods,

where the ªghting continued unabated.26 Hence instead of the violence burn-

ing out as Baghdad’s population unmixed, it simply moved as ascendant

Shiites attempted to conquer formerly Sunni territory. This produced a pattern

of continuing violence that was localized on the moving frontiers that sepa-

rated homogeneous neighborhoods. Moreover, these moving battle lines had

not exhausted the potential for bloodshed by mid- or even late 2007: Shiites

had conquered much but not all of Sunni Baghdad, leaving ample targets re-

maining for continuing predation when violence instead fell. This pattern of

sectarian killing offers little reason to believe that it ended of its own accord

by the middle of 2007, as the cleansing thesis argues.27

Testing the Surge 15

25. The following discussion is based entirely on original interviews with coalition ofªcers whoserved in Baghdad during the period in question.26. Of course, no human population is literally homogeneous in the sense that its makeup is abso-lutely uniform. There are always exceptions, and our usage is not meant to exclude this. By “ho-mogeneous,” we mean a substantial preponderance of one sect over another, following thecharacterizations provided by our interviewees and in demographic analyses such as the Gulf2000 Project.27. This does not constitute a general refutation of security dilemma theories of ethnosectarian vi-olence in Iraq or elsewhere. We do not claim that fear was unimportant in causing violence in Iraq.

In early to mid-2006, for example, the primary sectarian battleªelds in

the capital were in the neighborhoods bordering Khadamiya, just west of the

Tigris River to the city’s north, and around West Rashid in southern Baghdad

(see ªgure 1). The city’s pre-2006 sectarian demography was something of a

patchwork quilt, but west of the Tigris, central Baghdad was generally Sunni

with intermingled neighborhoods to the north and south and two predomi-

nantly Shiite enclaves beyond these: Khadamiya in the north and West Rashid

in the south.

International Security 37:1 16

What the evidence does establish, however, is that the pattern of violence offers no logical basis forIraq’s bloodshed to fall suddenly in mid-2007, whether the motive for that violence was fear orgreed. By mid-2007, conquest, not unmixing, was dominant in Baghdad’s sectarian warfare, andShiites’ conquest of the city was incomplete and apparently ongoing when the level of violencefell.

Figure 1. Baghdad Neighborhoods

This map was generated with TileMill software and OpenStreetMap shapefiles.

After the Samarra Mosque bombing in February, Muqtada al-Sadr’s Shiite

Jaish al-Mahdi (JAM) militia established lines of communication from its pri-

mary base in Sadr City, the large Shiite slum in northeast Baghdad, into beach-

heads west of the Tigris in these two Shiite quarters. In the north, the JAM

then began inªltrating the accessible mixed districts to the south and west of

Khadamiya. The result was a sharp increase in ªghting in Hurriyah, Shula,

and Washash. By October 2006 these cleansing efforts had succeeded, and

much of northern Baghdad had become homogeneously Shiite. Sectarian ªght-

ing did not stop, however—it simply moved. Instead of standing down with

the unmixing of intermingled Washash, the JAM drove south into the adjoin-

ing Sunni neighborhood of Mansour and southeast into predominantly Sunni

Karkh along Haifa Street. Instead of halting with the clearance of once-mixed

Shula and Hurriyah, the JAM pushed onward into Sunni Adel and southwest

into Ghazaliyah. In the process, they drove out residents from uniformly Sunni

apartment blocks for replacement with Shiite squatters, pushed the sectarian

frontier outward into homogeneously Sunni territory, and extended their line

of communications from Sadr City to enable further advances south and

southeast into the heart of Sunni central Baghdad.28

In southern Baghdad, the JAM moved outward from its base in West Rashid,

clearing areas with large Shiite populations such as Jihad, Bayaa, and Abu

T’Shir. Their offensive was not limited to rescuing Shiites, however. They also

attempted to expand outward into the predominantly Sunni neighborhoods of

Dora and Mechanic to the east, Ferat to the west, and Aamel and Sayidiyya to

the north.29

At no point, moreover, was the violence uniformly distributed over the in-

termingled sections of the city. Even during the ªghting for intermingled

Shula, Hurriyah, Washash, Sayidiyya, and Aamel, bloodshed was concen-

trated at the front lines of the JAM advance through these districts from their

bases in Khadamiya and West Rashid, with localities off these frontiers rela-

tively quiet. Nor was residence in a contiguous, homogeneously Sunni neigh-

borhood any guarantee of safety: the JAM offensive carried onward directly

into the Sunni heart of central Baghdad once it had cleared the way. In early

2006, the metropolitan districts of Karkh and Mansour comprised a homoge-

neously Sunni community of more than 1.5 million people over 10 to 20 con-

tiguous square miles; by mid-2007 perhaps half of this area had been cleansed

of its Sunni residents by the JAM and repopulated with Shiite squatters. The

Testing the Surge 17

28. U.S. Army Military History Institute, Iraq Surge Collection (henceforth MHI) audio ªles 10, 17,23, 35, and 55.29. MHI audio ªles 11, 18, 26, 36, 38, 55, and 61.

fact that the remainder was homogeneous and contiguous thus offered no rea-

son to expect that it would not be next on the list for conquest—the ªghting

had no more burned itself out in mid-2007 when the front line had advanced

partway into Sunni central Baghdad than it had in mid-2006 when it was ap-

proaching the boundaries of the once mixed-sect neighborhoods in the city’s

north and south. Something other than the natural completion of a process of

unmixing had to be at work for the violence to end when it did.30

The Awakening

Proponents of the Awakening thesis claim that violence declined in 2007 be-

cause the Sunni insurgency abandoned its erstwhile AQI allies in exchange for

U.S. payments of $300 per ªghter per month as “Sons of Iraq” (SOI) and a

promised cease-ªre. In this view, these actions yielded an uneasy truce in

which still-armed, unbeaten insurgent factions stopped ªghting for reasons

that had little or nothing to do with the surge.31

Supporters of the Awakening and the synergy schools both see Sunni re-

alignment as necessary for Iraq’s 2007 violence reduction; the former implies it

was also sufªcient for this. We evaluate the Awakening’s necessity below. In

this section, we consider its sufªciency: If there had been an Awakening but

without the reinforcements and new methods brought by the surge, would

Iraq’s violence still have fallen as broadly and rapidly as it did in 2007?

The experience of 2004–06 sheds light on this counterfactual. During this pe-

riod, Sunnis made at least four attempts to realign with coalition forces; none

succeeded. Each time, Sunni tribal leaders had become alienated by AQI’s bru-

tal methods and tried to break with AQI by negotiating local cease-ªres with

U.S. commanders. Each time, Sunni tribesmen agreed to defend their commu-

nities from al-Qaida in exchange for payments from the United States or the

Iraqi government. Yet none of these efforts received the kind of protection that

International Security 37:1 18

30. MHI audio ªles 11, 18, 38, and 55. Mansour’s and Karkh’s size and population were calculatedfrom LandScan 2008 data. Similar patterns characterized violence east of the Tigris. In Rusafa, forinstance, Sunnis lived mostly in mahala-sized clusters such as the roughly 3.5-square-kilometerSheikh Omar neighborhood. This was divided from the Shiite sections of al-Fadl to the south by amarket known as “Line Square.” The Sheikh Omar neighborhood’s perimeter was walled; LineSquare was defended on both sides with barriers and snipers; and this is where the sectarian vio-lence mainly occurred. MHI audio ªle 14. In the Madain district south of Baghdad, the sects werealso largely divided into homogeneous clusters, with Sunnis living to the north and Shiites to thesouth. Violence largely took place along this fault line, with JAM evicting Sunnis and emplacingsquatters as they left. MHI audio ªle 51.31. On the Awakening thesis, see Long, “The Anbar Awakening”; Simon, “The Price of theSurge”; Lynch, “Sunni World”; Michaels, A Chance in Hell; Green, “The Fallujah Awakening”;Couch, The Sheriff of Ramadi; Kober, “Did the Surge Work?”; Anderson, “Inside the Surge”; andRosen, “The Myth of the Surge.”

the surge offered to the Anbar Awakening, and without this protection, none

of these efforts proved able to survive and spread in the face of insurgent

counterattacks.32

The ªrst of these four failed Sunni realignments involved the Albu Nimr

tribe in 2004. Based in Anbar but with members living as far as Baghdad, the

Nimr were a cohesive tribe with a larger membership than the Albu Risha—

the tribe that ultimately catalyzed the late 2006 Awakening. The Nimr reached

out to U.S. forces in early 2004 to make common cause against al-Qaida by

standing up tribesmen as local police and civil defense forces in exchange for

U.S. money, weapons, and support. In 2004, however, the U.S. military had lit-

tle to offer in the way of direct protection; a single Special Forces detachment

of a dozen soldiers was assigned to work with the Nimr and coordinate their

security. Meanwhile, the limited conventional force presence in Anbar focused

its attention on the offensive in Fallujah, with no meaningful capacity to pro-

tect Nimr tribesmen elsewhere. Coalition requests to the Nimr to assist in

Fallujah produced tension when the Nimr demurred, and when AQI began se-

rious counterattacks against the Nimr in mid-2004, the coalition’s inability

to protect its allies became clear. Many Nimr tribesmen were killed, others

melted away, and the alliance collapsed.33

The second failed realignment came in the spring of 2005. Sunnis from the

Albu Mahal tribe in al-Qaim (together with Albu Nimr elements from the city

of Hit) created an armed resistance movement dubbed the “Hamza Brigade.”

AQI fought back, and by May the Hamza Brigade was seeking U.S. mili-

tary assistance. They received little. U.S. operations near al-Qaim (Operation

Matador) were not coordinated with the Hamza leadership, which complained

that the coalition was not protecting them from AQI attack. By September,

Hamza forces had been driven from al-Qaim; the Hamza Brigade had dis-

solved; and the tribesmen who continued to resist AQI had withdrawn to the

town of Aqashat and been marginalized.34

Testing the Surge 19

32. Secondary literature is cited where relevant, but the description of previous Awakening at-tempts is again solely based on original interviews.33. MHI audio ªle 48. See also Timothy S. McWilliams and Kurtis P. Wheeler, eds., Al-Anbar Awak-ening, Vol. 1: American Perspectives: U.S. Marines and Counterinsurgency in Iraq, 2004–2009(Quantico, Va.: Marine Corps University, 2009), pp. 54, 62; Steve Negus, “Home-Grown PoliceForce Takes On Iraq Insurgents,” Financial Times, March 31, 2006; and Nelson Hernandez, “IraqisBegin Duty with Refusal; Some Sunni Soldiers Say They Won’t Serve outside Home Areas,” Wash-ington Post, May 2, 2006.34. MHI audio ªle 48. See also Long, “The Anbar Awakening,” p. 78; James A. Russell, Innovation,Transformation, and War: Counterinsurgency Operations in Anbar and Ninewa Provinces, Iraq, 2005–2007 (Stanford, Calif.: Stanford University Press, 2011), p. 60; Gary W. Montgomery and Timothy S.McWilliams, eds., Al-Anbar Awakening, Vol. 2: Iraqi Perspectives: From Insurgency to Counterinsur-gency in Iraq, 2004–2009 (Quantico, Va.: Marine Corps University, 2009), p. 142; and Malkasian,

In the fall of 2005, the Desert Protectors, a militia organized by U.S. Special

Forces in conjunction with Operation Steel Curtain in al-Qaim, attempted a

third realignment. The U.S. commandos trained and equipped forty-ªve to

sixty Sunnis, many of whom were reportedly remnants of the Hamza Brigade,

and committed them in al-Qaim during the operation. The program was then

used to route local recruits into the Iraqi army and police. At its peak, the

Desert Protectors provided perhaps 1,000 Sunni recruits, but the organization

broke down after the coalition insisted that participants serve not as strict

home-defense forces and instead agree to be redeployed for service elsewhere.

In response, about one-third of the members resigned, and the program largely

disbanded.35

The fourth failed realignment was dubbed the “Anbar People’s Council”

and began in late 2005. Organized by seventeen tribal elders mostly from the

Fahad tribe, the Council was led by Muhammad Mahmud Latif al-Fahadawi

and Sheikh Nasser al-Fahadawi. Its leaders and many of its members were in-

surgents from the 1920s Brigade (a prominent Sunni guerilla faction) who had

become disaffected by AQI’s criminal activities and expropriation of local

smuggling income. On November 28, 2005, they decided to break with AQI

and support the coalition, directing tribesmen into the police for local security

duty. The coalition accepted these recruits, but failed to protect their leader-

ship. By early 2006, AQI counterattacks against the group had become ex-

tremely violent: al-Qaida bombed a police station during an Anbar Council

recruitment drive in January, killing seventy. Although initially resilient in

the face of this violence, the Council could not hold out indeªnitely: by

late January, AQI had killed almost half the founding elders, including Sheikh

Nasser. By the end of the month, the group had disbanded. The Anbar People’s

Council was notable for its similarity to the eventual Anbar Awakening move-

ment: it had a wide popular base (much wider than the Albu Risha tribe that

catalyzed the 2006 Awakening); it included a substantial number of disaffected

insurgents; and its leaders and foot soldiers accepted signiªcant personal risk

to combat al-Qaida. Yet its inability to defend itself from counterattack shut it

down within weeks.36

There is thus ample evidence of earlier attempts by Sunni tribes, including

International Security 37:1 20

“Did the Coalition Need More Forces in Iraq?” p. 123. Later, after U.S. Operations Steel Curtainand Steel Curtain II cleared al-Qaim in November 2005, U.S. forces reengaged the Albu Mahal inholding the area, but the Hamza Brigade had effectively been disestablished. MHI audio ªle 48.35. MHI audio ªles 29 and 48.36. MHI audio ªles 46 and 48. See also McWilliams and Wheeler, Al-Anbar Awakening, Vol. 1,pp. 114, 125–126; and Niel Smith and Sean MacFarland, “Anbar Awakens: The Tipping Point,”Military Review, Vol. 88, No. 2 (March/April 2008), pp. 41–52.

former insurgent groups, to realign against their erstwhile AQI allies prior to

the surge. Yet none succeeded. For some, unwillingness to deploy outside their

home districts contributed to their breakup. For at least three of the four cases,

however, the decisive cause of failure appears to have been their inability to

withstand counterattack.

This should not be surprising. All insurgencies face a constant risk of fac-

tionalism and defection, which can easily lead to annihilation by larger, better-

equipped state militaries.37 Self-preservation thus compels insurgents to put

down incipient defections with brutal violence lest the defection spread, and

AQI was unusually ruthless in this regard. Any Sunni tribe that broke with al-

Qaida could expect ªerce retaliation. In the successful 2006–07 Awakening, the

Sons of Iraq were under constant threat of reprisal. Interviewees reported in-

surgents assassinating SOI leaders or engaging SOI units in prolonged gun-

ªghts in Amiriyah, Awja, Babil, East Rashid, Haditha, Jisr Diyala, Khadamiya,

Kirkuk, Ramadi, Sadr al Yusuªyah, Tarmia, and Tikrit. Insurgents used impro-

vised explosive devices (IEDs) to target recruitment drives and to attack SOI

checkpoints, and the 2006–07 Awakening’s original leader, Sheikh Sattar Albu

Risha, was himself killed with an IED in 2007.38 The 2006–07 SOIs, however,

had the surge to protect them from these attacks; the prior attempts did not,

and none survived long enough to change the war in any fundamental way.

The fact that the 2006–07 Awakening received coalition protection that its

predecessors lacked was not an accident. The pre-surge U.S. military was in no

position to provide the security that Sunnis needed. This was partly because

smaller pre-surge U.S. forces had less capacity for protection,39 but it was also

because the prevailing pre-surge doctrine was ill suited to the job. With the ex-

ception of occasional experiments by innovative local commanders, pre-surge

methods normally emphasized force protection via mounted patrols; deploy-

ment in large, fortiªed bases; and operations in large formations without

sustained informal contact with the population. These dispositions delayed re-

sponsiveness in assisting Sunnis under attack, made it hard for tribesmen to

Testing the Surge 21

37. See, for example, Paul Staniland, “Between a Rock and a Hard Place: Insurgent Fratricide, Eth-nic Defection, and the Rise of Pro-State Paramilitaries,” Journal of Conºict Resolution, Vol. 56, No. 1(February 2012), pp. 16–40; and Stathis N. Kalyvas, The Logic of Violence in Civil War (New York:Cambridge University Press, 2006).38. See, for example, MHI audio ªles 4, 9, 15, 18, 23, 25, 35, 37, 40, 45, 48, 49, 50, and 69.39. Pre-surge U.S. troop strength averaged 135,000 to 140,000, and exceeded 145,000 in only sevenof the forty-three months between August 2003 and February 2007. Michael O’Hanlon and IanLivingston, “Iraq Index: Tracking Variables of Reconstruction & Security in Post-Saddam Iraq”(Washington, D.C.: Brookings Institution, August 2011), p. 13, http://www.brookings.edu/saban/iraq-index.aspx, accessed March 6, 2012. U.S. troop strength while the Anbar People’s Council andthe Desert Protectors were active averaged 150,000 to 155,000; neither of the other pre-surge re-alignment attempts occurred with more than 140,000 U.S. troops in Iraq.

communicate with U.S. troops without surveillance by their enemies, and

made it all but impossible to protect realigning tribes from inªltration by AQI

operatives or to protect their families from AQI assassination teams. In fact,

some local U.S. commanders did try to cooperate with realigning tribesmen

prior to the surge. The pre-2007 troop strength and doctrine, however, made

such cooperation difªcult even when both sides sought it. Without the surge

or its methods, realignment faced hurdles too high to overcome.40

Comparing Surge and Synergy

If the Anbar Awakening could not have survived without the surge, could the

surge have succeeded without the Awakening? The answer to this question is

what separates the surge and synergy theses, and we examine that question in

this section via a series of tests designed to distinguish these arguments from

each other.

logic of the surge thesis

The surge thesis has two main logical components.41 First, it holds that 30,000

U.S. reinforcements proved necessary to extend security over the critical sec-

tions of western and central Iraq. Coalition offensives had been able to clear

and even hold particular places at particular times prior to 2007; the problem,

in this view, was that insurgents had simply moved from the cleared areas to

others, leading to deterioration elsewhere in a balloon-squeezing phenomenon

that prevented security from improving overall. The surge, by contrast, is said

to have provided enough troops to clear and hold much wider expanses, pre-

venting the return of insurgents and sustaining security gains.42

Second, the surge brought the new tactics described above. Previous U.S.

International Security 37:1 22

40. Note that the fall 2006 Anbar realignment that initiated the Awakening occurred in Col. SeanMacFarland’s AO, where one of the occasional early experiments with Petraeus-like methods wasongoing. The surge brought such methods across the theater, and was thus instrumental in theAnbar Awakening’s ability to spread beyond its origin in this AO.41. The description below is drawn from Kagan, The Surge; McCain and Lieberman, “The SurgeWorked”; Boot, “The Truth about Iraq’s Casualty Count”; Crider, “A View from Inside the Surge”;Collier, “Now That We’re Leaving Iraq, What Did We Learn?”; Bull, “Mission Accomplished”; andAndrade, Surging South of Baghdad.42. Examples include Operations Iron Reaper and Iron Harvest, which we discuss in more detailbelow. Many analysts also emphasize concomitant improvements in Iraqi Security Forces (ISF),which they believe provided much of the surge’s effective strength. See, for example, Kagan, TheSurge, pp. 137–165. Here we treat the ISF buildup and U.S. reinforcements together as the surge,though Iraqi methods and skills lagged far behind the Americans’ throughout 2007. Note that thesurge’s main effort was initially in Baghdad with additional forces deployed to Anbar and Diyala.The resulting increase in troop density was felt more broadly over time, as the stabilization of ar-eas to which the surge brigades initially deployed enabled forces to be moved elsewhere.

methods under Gen. George Casey had emphasized transition to Iraqi security

forces. Implementation varied from place to place and commander to com-

mander, but many emphasized reducing U.S. vulnerability and visibility

among a resentful population until indigenous forces could take over. General

Petraeus, in this view, pushed U.S. forces out among the population and

tasked U.S. troops with protecting Iraqi civilians themselves. This was not en-

tirely unprecedented; innovative individuals such as Col. H.R. McMaster

at Tal Afar in 2005 and Col. Sean MacFarland in Ramadi in 2006 had experi-

mented with similar approaches on a local basis prior to the surge. What

changed in 2007 was that Petraeus insisted on their consistent, theaterwide

adoption and thus regularized such methods across Iraq.

The surge-only argument is thus more than just a claim about reinforce-

ments: it is centrally an argument about more troops and new doctrine for

their use. Of course, troop count and doctrine are logically independent. In

principle, the surge-only thesis could be sustained if one, the other, or the com-

bination proved decisive. What this school does require, however, is that some

combination of the two was both necessary and sufªcient to bring violence

down to something similar to the observed 2008 levels.

logic of the synergy thesis

The surge-Awakening synergy thesis, by contrast, sees the reinforcements and

doctrinal changes as necessary but insufªcient.43 In this view, the surge was

too small, and the impact of doctrinal changes insufªcient, to defeat a deter-

mined insurgency before the reinforcements’ time limit was reached and their

withdrawal began. Hence the surge without the Awakening would have

improved security temporarily but would not have broken the insurgency,

which would have survived and returned as the reinforcements went home.

The surge added a temporary, yearlong boost of about 30,000 U.S. troops to a

pre-surge coalition strength of about 155,000 foreign and 323,000 Iraqi troops

and police as of December 2006 (Iraqi Security Forces, or ISF, grew by about

another 37,000 by September 2007, when violence had begun to drop).44 Thus

the surge entailed only a marginal increase in troop density: an expansion of

less than 15 percent overall and perhaps 20 percent in U.S. strength. Half of the

overall increase, moreover, was in Iraqi forces, which were far from proªcient

in the new U.S. methods by 2006–07.45 And as mentioned above, the U.S. com-

Testing the Surge 23

43. The account below is drawn from Biddle, “Stabilizing Iraq from the Bottom Up”; Biddle,O’Hanlon, and Pollack, “How to Leave a Stable Iraq”; Kahl, “Walk before Running”; and Mal-kasian, “Did the Coalition Need More Forces in Iraq?”44. For troop counts, see O’Hanlon and Livingston, Iraq Index, pp. 13, 17.45. The Jones Commission, tasked by the U.S. Congress with assessing ISF capability and poten-

ponent had only about a year in which to function at this strength, after which

it was to return to pre-surge numbers or fewer. For this reinforcement per se to

have been decisive, one must assume that previous troop density lay just be-

low some critical threshold that happened to be within 20 percent of the pre-

surge value. Although this coincidence cannot be excluded, there is no prima

facie reason to expect it.46

For synergy proponents, the Awakening was thus necessary for the surge to

succeed. In this view, the Awakening had three central effects. First, it took

most of the Sunni insurgency off the battleªeld as an opponent, radically

weakening the enemy. Second, it provided crucial information on remaining

holdouts, and especially AQI, which greatly increased coalition combat effec-

tiveness. And third, these effects among Sunnis reshaped Shiite incentives,

leading their primary militias to stand down in turn.

As for the ªrst two points, although the SOI movement never comprised just

former insurgents, the insurgency nevertheless provided much of the SOIs’

combatant strength—and the bulk of the secular Sunni insurgency nationwide

became SOIs over the course of 2007. By the end of the year, SOI strength

nationwide had reached 100,000 members, under more than 200 separate con-

tracts. As insurgents progressively realigned in this way, the remaining insur-

gency shrank dramatically. The fact that so many SOIs were former insurgents

also made the SOIs uniquely valuable coalition allies: they knew their erst-

while associates’ identities, methods, and whereabouts in ways that govern-

ment counterinsurgents rarely do. When insurgents who had been allied

with AQI realigned as Sons of Iraq, the coalition suddenly gained intelligence

International Security 37:1 24

tial, put it this way in September 2007: “The challenge for the [Iraqi] Army is its limited opera-tional effectiveness, caused primarily by deªciencies in leadership, lack of disciplinary standards,and logistics shortfalls.” Iraqi police rated even harsher assessments: “In general, the Iraqi PoliceService is incapable today of providing security at a level sufªcient to protect Iraqi neighborhoodsfrom insurgents and sectarian violence. . . . The National Police have proven operationally ineffec-tive, and sectarianism in these units may fundamentally undermine their ability to provide secu-rity. The force is not viable in its current form.” The Commission saw the ISF as improving, but farfrom effective, in 2007. Gen. James L. Jones, U.S. Marine Corps, chairman, The Report of the Inde-pendent Commission on the Security Forces of Iraq (Washington, D.C.: Center for Strategic and Inter-national Studies, September 6, 2007), pp. 9–10.46. This is especially true given that the U.S. military fell well short of its own doctrinal standardthat successful counterinsurgents require 20 troops per 1,000 civilians to be protected. Headquar-ters, Department of the Army, FM 3-24: Counterinsurgency (Washington, D.C.: U.S. GovernmentPrinting Ofªce, December 2006), par. 1-67. Iraq’s population is roughly 30 million. Central Intelli-gence Agency, The World Factbook: Iraq, https://www.cia.gov/library/publications/the-world-factbook/geos/iz.html, accessed March 30, 2012. FM 3-24 thus implies a need for about 600,000soldiers in Iraq—and, as discussed earlier, it is unclear how much the ISF contributed to the effort.The utility of FM 3-24’s standard is questionable (see Friedman, “Manpower and Counterinsur-gency”), but even those who accept it would have little reason to expect that the surge bumped co-alition troop density above some critical threshold in Iraq.

on AQI membership, cell structure, the identity of safe houses and bomb-

making workshops, and locations of roadside bombs and booby traps. Guer-

rillas rely on stealth and secrecy to survive against heavily armed government

soldiers. When SOIs lifted this veil of secrecy, coalition ªrepower guided by

SOI intelligence became extremely lethal, creating ever-increasing incentives

for holdouts to seek similar deals for themselves; soon only committed AQI fa-

natics remained, marginalized in a few districts in Iraq’s northwest.47

In the synergy account, Sunni realignment in turn had major consequences

for Shiite militias such as the Jaish al-Mahdi.48 Many of these militias began as

self-defense mechanisms to protect Shiite civilians from Sunni attack, but they

grew increasingly predatory as they realized they could exploit a dependent

population. Rising criminality in turn created ªssiparous tendencies as fac-

tions with their own income grew increasingly independent of their leader-

ship. When the SOIs began appearing, the Sunni threat waned, and with it the

need for defenders. At the same time, the SOI cease-ªres freed arriving U.S.

surge brigades to focus on Shiite militiamen. These developments created mul-

tiple perils for militia leadership. In previous ªreªghts with U.S. forces, the

JAM in particular had sustained heavy losses but easily made them up with

new recruits given its popularity. Shiites’ growing disaffection with militia

predation, however, coupled with declining fear of Sunni attack, threatened

leaders’ ability to make up losses with new recruits. At the same time, intra-

Shiite violence among rival militias, especially between the Badr Brigade and

the JAM, posed a rising threat from a different direction. When Shiites were

uniªed by a mortal Sunni threat and U.S. forces were tied down by insur-

gents and AQI, these internal problems were manageable. But as the Sunni

threat waned, Shiite support weakened, internal divisions multiplied, and

U.S. troop strength grew, Shiite militias’ ability to survive new battles with co-

alition forces fell. In the synergy account, these challenges persuaded Muqtada

Testing the Surge 25

47. Most interviewees with ªrsthand knowledge of SOIs reported that these units contained for-mer insurgents, not just in Anbar but across central Iraq, including Al Dur, Awja, Baladrooz,Habbaniyah, Hit, Kirkuk, Narwan, Northern Babil, Northwest Wasit, Rawah, Salman Pak, SouthDiyala, Tikrit, and Wynot, and the Baghdad neighborhoods of al-Rusafa, Amiriyah, Dora, EastRashid, Khadamiya, Madain, Mahmoudiya, Sayidiyya, and Tarmia. MHI audio ªles 4, 7–11, 13, 18,20, 23–26, 30, 35, 41, 43, 49, 51, 53, 54, 63, 64, 68, and 69.48. The discussion here follows the synergy literature; it focuses on the JAM as the most importantof the Shiite militias and one whose military incentives mirrored those of its rivals. On other Shiitemilitias, see Michael Harari, Status Update: Shi’a Militias in Iraq (Washington, D.C.: Institute for theStudy of War, 2010); Anthony Cordesman, The Iraqi Insurgency and the Risk of Civil War: Who Are thePlayers? (Washington, D.C.: Center for Strategic and International Studies, 2006); Patrick Gaughen,“Backgrounder #17: The Fight for Diwaniyah: The Sadrist Trend and the ISCI Struggle for Suprem-acy” (Washington, D.C.: Institute for the Study of War, 2007); and Anthony Shadid, “In a Landwithout Order, Punishment Is Power: Conºicts among Shiites Challenge a Village Sheikh in South-ern Iraq,” Washington Post, October 22, 2006.

al-Sadr to stand down rather than risk another beating from the coalition, and

the result was his announced cease-ªre of August 2007—which took the pri-

mary Shiite militia off the battleªeld, leaving all of 2006’s major militant

groups under cease-ªres, save a marginalized remnant of AQI, and producing

the radical violence reduction of late 2007 and thereafter.49

Proponents of the synergy thesis thus see the Awakening as necessary

for the surge to succeed. In this view, however, neither the surge nor the

Awakening was sufªcient, nor did these factors combine in an additive way.

As noted above, Sunni groups had attempted similar realignments on pre-

vious occasions—and those earlier attempts had all failed at great cost. For the

synergy school, what distinguished the failures from the successful 2007

Awakening was a coalition force that could protect insurgent defectors from

counterattack. The surge may not have been large enough to suffocate a deter-

mined insurgency, but it was large enough to enable cooperation with turncoat

Sunnis and exploit their knowledge to direct coalition firepower against the

still-active insurgents, enabling them to survive the kind of retaliation that had

crippled their predecessors.

This U.S. contribution required the surge’s doctrinal element as well as its

reinforcements. After all, the initial experiments that became the Awakening

and the SOI movement predated the reinforcements: it was in Anbar in the fall

of 2006 that Sheikh Sattar and his Albu Risha tribe ªrst worked out an arrange-

ment to assist U.S. forces under Colonel MacFarland in exchange for physical

protection against counterattack. As Colonel McMaster had done in Tal Afar

and others had tried elsewhere, MacFarland had anticipated the methods that

Petraeus would shortly institute across the theater. Those methods were

necessary for the delicate process of establishing trust between mutually wary

parties and enabling the necessary speed and effectiveness of security coopera-

tion. The fall 2006 Awakening in Anbar could survive because it was sup-

ported with the new methods, and it could spread across most of Iraq in less

than a year because the surge spread those methods across a reinforced the-

ater. For the synergy school, both the Awakening and the surge were therefore

necessary, as each reinforced the other in close interaction; and the surge’s doc-

trinal component, not just its numerical reinforcement, was necessary.

International Security 37:1 26

49. An exception was the post-2007 combat between Sadr’s JAM and coalition forces during theIraqi government’s spring 2008 “Charge of the Knights” offensive in Basra and the follow-on oper-ations in Amarah and Sadr City. This offensive led to a brief spike in violence, which quickly sub-sided when Sadr again stood down in a cease-ªre negotiated with Nouri al-Maliki’s governmentafter the JAM proved unable to hold; the 2008 cease-ªre’s logic was similar to its 2007 predeces-sor’s, and the 2008 violence did not produce more than a temporary exception to the trend of radi-cal violence reduction after 2007. On the Charge of the Knights offensive and ensuing combat, seeMarisa Cochrane, The Battle for Basra, Iraq Report, No. 9 (Washington, D.C.: Institute for the Studyof War, June 23, 2008).

evidence: surge versus synergy

Which, then, is the stronger explanation, surge or synergy? The critical distinc-

tion concerns the SOIs. Both schools’ proponents see the surge as necessary;

only synergy thesis adherents see the SOIs as essential. If the surge-only view

is correct, the SOIs were thus either inconsequential or epiphenomenal. Re-

garding the latter, even a crucial Awakening could still be consistent with a

surge-only explanation if the surge created it. Of course, Sheikh Sattar’s origi-

nal Awakening predated the surge, so it could not have been caused by it.

The vast majority of the SOIs, however, came later and elsewhere, after surge

brigades had begun to deploy—the transition from a small-scale experiment

in one part of Iraq to a widespread movement that ªelded 100,000 members

took place almost entirely after the surge was well under way. If the surge

was as powerful as is often claimed, rational insurgents would have had

strong reasons to abandon the ªght, and the SOI program offered an ideal

vehicle.

To assess these possibilities, the critical observable implications concern the

trajectory of the reduction in violence and its relationship to the timing of SOI

standup. The surge-only thesis implies that for any given area of operations,

combat before SOI standup should have been sufªcient to reverse the 2006

trend of increasing bloodshed and to put local violence on a downward trajec-

tory steep enough to pacify the AO before the surge ended and reinforcements

were withdrawn. Where this is so, subsequent SOI standups would be super-

ºuous, either because they were unnecessary or because they would be indis-

tinguishable from a rational decision by insurgents to stop contesting a surge

that was beating them anyway. In such cases, the surge would have sufªced

without the SOIs, whether they eventually appeared or not, and their appear-

ance would be epiphenomenal to the surge. By contrast, the synergy thesis im-

plies that for any given AO, the reduction in violence before SOI standup

should be too slow to pacify the AO before the surge’s end. The reduction in

violence afterward, however, should accelerate to a pace that could reduce

bloodshed to roughly 2008 levels by the surge’s end, and by the end of U.S. po-

litical patience with such intense combat in Iraq—which had neared the break-

ing point even by mid-2007 and would surely not have survived into the 2008

presidential campaign season. If so, the surge would be necessary but insuf-

ªcient, and unable to pacify Iraq by itself or motivate rational Sunnis to

realign.

To test these implications, we compared SIGACTs trends before and after

SOI standup in each of the 38 AOs for which our interviews provide speciªc

standup dates.50 We computed violence slopes using ordinary least squares re-

Testing the Surge 27

50. In the two AOs where interviewees rotated out before the SOIs stood up, we used Iraq Recon-

gression on the three months of combat preceding SOI standup in that AO,

and the three months afterward.51 Figure 2 illustrates the results by showing

trends in SIGACTs over time for each AO, with SOI standup dates superim-

posed. Table 1 provides descriptive information on each AO, including slope

estimates and other details.

The results suggest that SOIs played a crucial role in reducing Iraq’s vio-

lence in 2007. As table 2 shows, 24 of 38 AOs where SOIs stood up (63 percent)

show violence trending downward more sharply after SOI standup than be-

fore. The difference, moreover, is large: across all 38 AOs, the average rate of

reduction before SOI standup was 2.5 percentage points per month; the rate af-

ter standup was 5.8 percentage points per month, or roughly two and a half

times greater.52 Of course, these are aggregate statistics. Table 2 thus breaks the

data down into subsets by timing of SOI standup, location of AOs, and vio-

lence patterns therein.

The results in table 2 suggest that SOIs had bigger effects in more important

AOs. For example, in AOs where SOIs stood up prior to August 2007 (when

ªghting was generally heaviest), violence declined faster after SOIs stood up

in 78 percent of cases. Whereas violence was increasing in each of these AOs at

standup, it reversed and plummeted thereafter, falling by more than 8 percent-

age points per month on average.53 For AOs in Baghdad and Anbar (the

International Security 37:1 28

struction Management System data to identify the date of ªrst payments to SOIs in that AO. For adescription of these data, see Berman, Shapiro, and Felter, “Can Hearts and Minds Be Bought?”51. The independent variable for these regressions is time; the dependent variable is monthlySIGACTs as a percentage of the maximum value that AO experienced from 2004 to 2008. This nor-malization facilitates cross-AO comparisons, because a drop of 10 SIGACTs per month could be amajor change in a quiet AO, but a marginal change elsewhere. Thus, an estimated coefªcient of�0.06 would show violence declining in that AO by six percentage points per month. Two AOsproduced equal coefªcients before and after SOI standup; these ties were broken by examiningone- and two-month intervals (see table 1). All regressions and supporting data are available athttp://esoc.princeton.edu.52. These ªndings are robust to a variety of alternative speciªcations. When we examine other in-tervals of equal length before and after SOI standup, the SOIs correlate with faster rates of violencereduction. If we were to shorten the intervals to two months, for example, the SOIs’ apparent im-pact would increase, with violence declining on average by 1.2 percentage points per month priorto SOI standup and by 5.8 percentage points thereafter. There is no interval between one andtwelve months for which the violence reduction rate does not increase by at least a factor of 1.9 af-ter SOIs stand up. Nor do the intervals need to be symmetric to support synergy: when the slopeof violence is computed for any interval from one to twelve months after SOI standup, it fallsfaster on average across all 38 AOs than it does for any interval from one to twelve months prior toSOI standup. We also examined the robustness of these results by dropping all SIGACTs that werepositively identiªed as not occurring from combat. All of these patterns remained substantivelythe same: for example, violence fell by 6.2 percentage points per month on average in the threemonths following standup, versus 2.5 in the three months prior. Across a range of intervals andways of measuring insurgent attacks, violence thus drops faster after SOI standup.53. Table 2 also demonstrates that, for AOs where SOIs stood up during 2007, 69 percent supportthe synergy thesis, with a ªvefold acceleration in violence reduction.

Testing the Surge 29

Fig

ure

2.

Vio

len

ce

Tre

nd

sb

yA

rea

of

Op

era

tio

n

NO

TES:

Are

aof

opera

tion

boundaries

and

Sons

of

Iraq

(SO

I)sta

ndup

date

sfr

om

auth

orin

terv

iew

s.D

ashed

line

show

sm

onth

of

firs

tSO

Ista

ndup.

SIG

AC

T(s

ignific

ant

activitie

s)

counts

based

on

decla

ssifie

ddata

from

Multin

ational

Forc

es–Iraq.

Tab

le1.

Are

as

of

Op

era

tio

nw

ith

So

ns

of

Iraq

(SO

I)S

tan

du

p

Are

ao

fO

pera

tio

nLo

cati

on

Peak

Vio

len

ce

SO

IS

tan

du

pD

ate

Pre

-SO

IV

iole

nce

Tre

nd

Po

st-

SO

IV

iole

nce

Tre

nd

Co

nª

rmati

on

for

Syn

erg

y?

Al

Du

rS

ala

din

Pro

vin

ce

30

May07

�11.7

�11.7

yes

Al

Hilla

hB

ab

ilP

rovin

ce

18

Au

g07

�8.3

�11.1

yes

Am

iriy

ah

West

Bag

hd

ad

85

May07

�15.4

�24.5

yes

Ara

bJab

ou

rS

ou

thB

ag

hd

ad

83

Au

g07

�3.9

�2.0

yes

Ash

Sh

arq

at

Sala

din

Pro

vin

ce

3A

ug

07

�16.7

�20.0

no

AS

RG

old

en

So

uth

of

Sam

arr

a150

Feb

08

�3.6

�0.8

no

Aw

jaS

ou

tho

fTik

rit

82

May07

�3.7

�5.5

yes

Bala

dro

oz

Diy

ala

Pro

vin

ce

42

May08

�1.2

�3.6

yes

Bayji

Sala

din

Pro

vin

ce

149

No

v07

�1.5

�3.3

yes

Do

raS

ou

thB

ag

hd

ad

264

Sep

07

�12.7

�16.1

yes

Fallu

jah

An

bar

Pro

vin

ce

148

May07

�4.1

�7.2

yes

FO

BK

als

uA

rea

Su

nn

iTri

an

gle

183

Ju

n07

�9.1

�12.6

yes

Gh

aza

liyah

No

rth

west

Bag

hd

ad

157

Ju

n07

�11.5

�4.7

no

Haq

lan

iya

So

uth

of

Had

ith

a160

Jan

07

�24.6

�7.2

no

Haw

ijah

Tam

eem

Pro

vin

ce

116

Jan

08

�14.8

�1.5

no

Hu

rriy

ah

No

rth

west

Bag

hd

ad

163

May07

�7.1

�4.0

yes

Jis

rD

iyala

Diy

ala

Pro

vin

ce

214

Ju

l07

�14.7

�20.8

yes

Kan

an

Diy

ala

Pro

vin

ce

21

Mar0

8�

3.3

�7.6

yes

Kark

hW

est

Bag

hd

ad

104

Oct0

7�

0.6

�1.7

yes

Kata

na

Do

wn

tow

nR

am

ad

i119

Oct0

6�

3.1

�11.6

yes

Kh

ad

am

iya

Urb

an

No

rth

west

Bag

hd

ad

71

May07

�16.8

�10.6

yes

Kh

alid

iyah

An

bar

Pro

vin

ce

47

Dec06

�3.4

�12.8

yes

Kh

an

Ban

iS

a’a

dD

iyala

Pro

vin

ce

43

Dec07

�9.8

�7.7

no

Lati

ªyah

So

uth

west

Bag

hd

ad

38

Dec07

�2.4

�2.4

yes

*

Man

so

ur

West

Bag

hd

ad

33

May07

�13.9

�8.5

yes

Man

su

riyat

al

Jab

al

Diy

ala

Pro

vin

ce

18

Jan

08

�27.8

�2.8

no

Mech

an

icS

ou

thB

ag

hd

ad

107

Dec07

�34.3

�1.2

no

Mu

qd

ad

iyah

Su

nn

iTri

an

gle

62

Feb

08

�18.1

�3.5

no

Rad

wan

iyh

aS

ou

thw

est

Bag

hd

ad

22

Ju

n07

�20.5

�5.9

no

Raw

ah

Weste

rnA

nb

ar

20

May07

�3.0

�7.5

yes

Ru

safa

Sh

eik

hO

mar

East

Bag

hd

ad

54

Ap

r08

�1.1

�1.1

yes

†

Sad

ral

Yu

su

ªyah

Su

nn

iTri

an

gle

82

Ju

n07

�1.3

�5.7

yes

Salm

an

Pak

So

uth

Bag

hd

ad

49

Au

g07

�6.5

�6.9

no

Sayid

iyya

So

uth

west

Bag

hd

ad

75

Au

g07

�3.7

�11.3

yes

Sh

eik

hH

am

ad

Villa

ge

Su

nn

iTri

an

gle

29

Sep

07

�9.3

�4.8

no

So

uth

ern

Tam

eem

So

uth

of

Kir

ku

k101

May08

�1.0

�1.2

no

Taji

No

rth

Bag

hd

ad

167

Ap

r07

�8.2

�11.4

no

Tam

imA

nb

ar

Pro

vin

ce

76

Jan

07

�13.3

�27.9

yes

Avera

ge

(N�

38)

91

�2.5

�5.8

“P

eak

Vio

len

ce”

isth

em

axim

um

nu

mb

er

of

sig

niª

can

tacti

vit

ies

ineach

are

ao

fo

pera

tio

nfo

ran

ym

on

thb

etw

een

2004

an

d2008.

“P

re-”

an

d“P

ost-

SO

IV

io-

len

ce

Tre

nd

s”

are

slo

pe

esti

mate

so

fp

iecew

ise

lin

ear

reg

ressio

ns

ªt

toth

eth

ree-m

on

thin

terv

alp

rio

rto

SO

Ista

nd

up

an

dth

eth

ree-m

on

thin

terv

alaft

erw

ard

.

*In

bo

tho

ne-

an

dtw

o-m

on

thin

terv

al

co

mp

ari

so

ns,

vio

len

ce

declin

ed

mo

rera

pid

lyaft

er

SO

Ista

nd

up

.†In

bo

tho

ne-

an

dtw

o-m

on

thin

terv

al

co

mp

ari

so

ns,

vio

len

ce

rose

mo

reslo

wly

aft

er

SO

Ista

nd

up

.

surge’s main effort), 71 percent are conªrmatory, and the reduction in violence

accelerated by nearly a factor of four after SOIs stood up. For AOs with at least

the median population or population density, 74 and 79 percent of AOs sup-

port synergy, respectively, with violence dropping an additional 6 to 8 percent-

age points per month after standup. For AOs where SOIs stood up while

violence was still above 50 percent of peak (i.e., where violence still had the

farthest to fall), the results are especially stark: 93 percent show improved

trends after standup, with the violence trends reversing from an average 7 per-

cent per month increase before standup to an 11 percent per month decrease

afterward.54 For fully half the entire sample (19 cases of 38), violence was

still rising when SOIs stood up; in every instance, violence trends reversed

thereafter.

International Security 37:1 32

54. Table 2 demonstrates that results also become more conªrmatory when one examines the19 AOs where violence was highest in the month of SOI standup.

Table 2. Comparisons of Violence Trends Pre- and Post-Sons of Iraq (SOI) Standup

Subset NAreas of OperationConªrming Synergy

Average Slopeprior to Standup

Average Slopeafter Standup

All Areas of Operation 38 24 (63%) �2.5 �5.8

SOI standup timingprior to August 2007 18 14 (78%) �1.5 �8.6during 2007 29 20 (69%) �1.2 �6.3

Region/demographyBaghdad/Anbar 24 17 (71%) �1.9 �7.0high population 19 14 (74%) �0.8 �7.2high population density 19 15 (79%) �0.5 �6.5

Violence Levels� 50 percent peak 15 14 (93%) �6.9 �10.9� median 19 14 (74%) �0.0 �8.5violence rising 19 19 (100%) �7.1 �9.8

“Areas of Operation Conªrming Synergy” are those where violence declines faster in thethree-month interval following SOI standup than the three-month interval beforehand.“Average Slope prior to Standup” and “Average Slope after Standup” give average vio-lence trends for all AOs meeting the given criterion.

The rows in the table present different subsets of the data: “high population” are AOswhose total population was above the median, and “high population density” are AOswhose population density was above the median relative to all 38 AOs. “� 50 percentpeak” are AOs where SOIs stood up in a month where violence was at least 50 percent ofits 2004–08 monthly maximum; “� median” are AOs where violence in the month whereSOIs stood up was greater than the median for all 38 AOs; and “violence rising” repre-sents AOs where SOIs stood up while violence was increasing over the previous three-month interval.

As noted above, interview evidence also supports crucial elements of the

synergy causal mechanism. Many SOI members were in fact former insur-

gents. These former insurgents did indeed provide important intelligence and

other support to U.S. forces; SOIs did indeed suffer frequent counterattacks

from holdouts and especially AQI; and U.S. forces often were required to come

to their defense when this happened, as synergy advocates claim.55

These ªndings suggest that if no SOIs had stood up, and if the pre-SOI

violence reduction rate seen in the AOs studied here had persisted for the

duration of the surge, then violence might have declined so slowly that Iraq—

especially Iraq’s key terrain—would have been far from stabilized when the

surge ended.56 The ªrst surge brigade deployed in February 2007; the last surge

brigade withdrew in July 2008.57 If violence had declined only at a rate of two

percentage points per month throughout this period (as seen, on average, prior

to SOI standup), then violence when the surge ended would have been no

lower than in mid-2006, and this after another ten months of intense combat

not seen in the historical case. Without SOIs, the data suggest that it could

have taken more than three years of grinding warfare with surge-scale troop

levels to bring the violence down to the levels achieved in a few months with

the SOIs; without the Awakening, violence would have remained very high

for a very long time—and certainly long after the surge brigades had gone

home.

Moreover, many of the apparently disconªrmatory AOs either pose anoma-

lies for the surge-only thesis, too, or offer important if partial support for ele-

ments of the synergy thesis all the same, or both. For example, in three of the

fourteen apparently disconªrmatory AOs (Ash Sharqat, Salman Pak, and Taji),

Testing the Surge 33

55. See MHI audio ªles 4, 9, 15, 18, 23, 25, 35, 37, 40, 45, 48, 49, 50, and 69.56. Note that casualty rates in COIN often increase following reinforcements, then decline thereaf-ter; this was so in Iraq in 2007, where U.S. casualties peaked three months into the surge in May2007. “Operation Iraqi Freedom,” iCasualties.org. In principle, this “darkest before the dawn” phe-nomenon could bias post-SOI violence reduction rates downward: if the pre-SOI slope calculationstraddled a crest in violence, this would artiªcially increase the apparent synergy conªrmationrate, but this did not happen here. In only 2 AOs (Sayidiyya and Radwaniyha) did violence peakwithin the three-month window prior to SOI standup, and only the Sayidiyya AO was otherwiseconªrmatory (and thus subject to potential conªrmation bias). Moreover, per note 51, SOIs corre-late with accelerated violence reduction no matter what intervals we